CN117670892A - Aquatic bird density estimation method and device, computer equipment and storage medium - Google Patents
Aquatic bird density estimation method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN117670892A CN117670892A CN202311668122.3A CN202311668122A CN117670892A CN 117670892 A CN117670892 A CN 117670892A CN 202311668122 A CN202311668122 A CN 202311668122A CN 117670892 A CN117670892 A CN 117670892A
- Authority
- CN
- China
- Prior art keywords
- bird
- image
- density
- panoramic image
- panoramic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000003860 storage Methods 0.000 title claims abstract description 13
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 20
- 238000003062 neural network model Methods 0.000 claims abstract description 14
- 230000011218 segmentation Effects 0.000 claims abstract description 11
- 241000271566 Aves Species 0.000 claims description 34
- 238000004422 calculation algorithm Methods 0.000 claims description 26
- 238000000605 extraction Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 8
- 238000002372 labelling Methods 0.000 claims description 6
- 238000005520 cutting process Methods 0.000 claims description 5
- 241000272517 Anseriformes Species 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 abstract description 20
- 238000013528 artificial neural network Methods 0.000 abstract description 12
- 238000005516 engineering process Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 9
- 230000010339 dilation Effects 0.000 description 6
- 238000009826 distribution Methods 0.000 description 6
- 238000011835 investigation Methods 0.000 description 6
- 238000011176 pooling Methods 0.000 description 6
- 241001465754 Metazoa Species 0.000 description 5
- 230000008859 change Effects 0.000 description 5
- 241000894007 species Species 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000012806 monitoring device Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 241000269799 Perca fluviatilis Species 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 230000033001 locomotion Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 208000028804 PERCHING syndrome Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention is applicable to the field of wetland protection, and provides a water bird density estimation method, a device, computer equipment and a storage medium, wherein the method comprises the following steps: acquiring a bird panoramic image; performing segmentation pretreatment on the bird panoramic image to obtain a plurality of undetermined images with the same size; estimating bird density of the undetermined image according to the depth density estimation neural network model to obtain a bird density map; and splicing the bird density map into a panoramic bird density map according to a splicing function, and calculating the number of birds. By utilizing an image flow technology and a depth neural network, the number and the density of the water birds can be accurately identified, counted and estimated, the limit of traditional water bird monitoring is broken through, and a more accurate and efficient water bird number estimation and density map generation method is provided.
Description
Technical Field
The application belongs to the field of wetland protection, and particularly relates to a waterfowl density estimation method, a waterfowl density estimation device, computer equipment and a storage medium.
Background
The perfected protection policies and effective practices of wild animals and habitats depend on timely and reliable species monitoring data, which is one of the important responsibilities of personnel in the protected area, and real-time video monitoring devices have many advantages in assisting species investigation, such as remote and non-invasive observation, real-time monitoring, cloud storage and recycling convenience, and the high-definition monitoring devices installed in the protected area provide many convenience and advantages for finding new species, monitoring the shapes of pending species, investigating animal activity hotspots and habitat usage models thereof, preventing illegal transactions, improving public awareness, and in the past decade, more and more of the protected area managers around the world are installing high-definition video monitoring devices in order to reduce the cost of observation, so as to assist in daily management of the protected area.
Traditional ground investigation is an important means for species monitoring, is difficult, time-consuming and labor-consuming depending on personnel investigation, can deviate data results, and greatly reduces monitoring cost and improves coverage expansion efficiency and accuracy, thereby providing new opportunities for local, regional and global protection research along with development of emerging technologies such as airborne and empty images, biological telemetry, infrared cameras, real-time video monitoring, passive sound recording and the like.
However, for a dead place where animals inhabit briefly, a manual investigation method for estimating the number and distribution of water birds by using a traditional spot counting or spline investigation is inaccurate in estimation, has few data samples, and has the problem that monitoring original data is easy to lose.
Disclosure of Invention
The embodiment of the application aims to provide a water bird density estimation method, which aims to solve the problems that the traditional manual investigation method is inaccurate in estimation, few in data sample and easy to lose in monitoring original data.
The embodiment of the application is realized in such a way that a water bird density estimation method comprises the following steps:
acquiring a bird panoramic image;
performing segmentation pretreatment on the bird panoramic image to obtain a plurality of undetermined images with the same size;
estimating bird density of the undetermined image according to the depth density estimation neural network model to obtain a bird density map;
and splicing the bird density map into a panoramic bird density map according to a splicing function, and calculating the number of birds.
Another object of an embodiment of the present application is a water bird density estimation device, the device comprising:
the image processing module is used for acquiring the bird panoramic image; performing segmentation pretreatment on the bird panoramic image to obtain a plurality of undetermined images with the same size;
the bird density estimation module is used for estimating the bird density of the undetermined image according to the depth density estimation neural network model to obtain a bird density map;
and the splicing module is used for splicing the bird density images into a panoramic bird density image according to a splicing function and calculating the number of birds.
Another object of an embodiment of the present application is a computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the method for estimating bird density.
Another object of an embodiment of the present application is a computer readable storage medium, on which a computer program is stored, which when executed by a processor causes the processor to perform the steps of the method for estimating bird density.
The aquatic bird density estimation method provided by the embodiment of the application obtains the aquatic bird information image in the whole monitoring scene by acquiring the panoramic image in the monitoring video, ensures the accuracy and reliability of the subsequent analysis result, cuts the image to the same size by the segmentation pretreatment of the panoramic image, ensures the calculation efficiency of the algorithm, retains the valuable characteristic information in the image as much as possible, improves the marking efficiency, reduces the marking cost, improves the convergence speed of network training by providing a depth density estimation neural network model, has higher accuracy than the density estimation of other similar algorithms, has lower calculation cost, splices the generated density image into the panoramic bird image of the whole video monitoring area by a splicing function, counts the number of birds, can reveal the perching rule in the bird migration process, especially the relationship between the environmental change and the change of the number of bird population,
the number and density of the waterbirds can be accurately identified, counted and estimated by using an image flow technology and a deep neural network. The method provides abundant data support for ecological research and wetland protection, reveals the distribution and utilization rules of the water birds in the wetland, effectively solves the detection, identification and counting problems in the traditional water bird monitoring method, breaks through the limitation of the traditional water bird monitoring, and provides a more accurate and efficient water bird quantity estimation and density map generation method.
Drawings
FIG. 1 is a flow chart of a method for estimating bird density according to an embodiment of the present disclosure;
fig. 2 is a flowchart of obtaining a panoramic image in a method for estimating a bird density according to an embodiment of the present application;
FIG. 3 is a flowchart of segmentation preprocessing in a method for estimating bird density according to an embodiment of the present application;
FIG. 4 is a flow chart of neural network construction in a method for estimating bird density according to an embodiment of the present application;
FIG. 5 is an image annotation diagram in a method for estimating bird density according to an embodiment of the present application;
fig. 6 is a schematic diagram of a neural network in a method for estimating bird density according to an embodiment of the present application;
FIG. 7 is an output illustration of a method for estimating bird density according to an embodiment of the present application;
FIG. 8 is a table comparing the estimation results of a method for estimating the density of a bird according to the embodiment of the present application;
FIG. 9 is a block diagram illustrating a structure of an xx device provided by an embodiment of the present application;
FIG. 10 is a block diagram of the internal architecture of a computer device in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
It will be understood that the terms "first," "second," and the like, as used herein, may be used to describe various elements, but these elements are not limited by these terms unless otherwise specified. These terms are only used to distinguish one element from another element. For example, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script, without departing from the scope of the present application.
As shown in fig. 1, in one embodiment, a method for estimating the density of a water bird is provided, which specifically includes the following steps:
step S102, acquiring a bird panoramic image.
In this embodiment, the panoramic image may capture image information in the whole scene, adopt the ecological law when panoramic image can better record birds perch, the relation between birds quantity and the environment, bird panoramic image can be the scene and take a photograph in real time and obtain, also can be the image in the image database, image acquisition equipment can professional surveillance camera head, cell-phone, satellite map etc., exemplary, panoramic image's acquisition place is the wetland, this wetland has the perch of four kinds of microtopography: low dykes and dams, shallow grassland, shallow water district and deep water district can effectually record the birds information in all regions through the album image, adopts spherical monitoring camera, can visit and watch through the IP address in real time based on the ONVIF protocol, and camera apart from ground 1.70 meters fixed mounting, and the biggest scope of zooming is 60 times, can shoot clear birds image. The camera can horizontally rotate 360 degrees, and the vertical rotation range is-45 degrees to 90 degrees. The camera may operate between-40 deg. and 40 deg. or in a humidity environment up to 93%.
And step S104, carrying out segmentation pretreatment on the bird panoramic image to obtain a plurality of undetermined images with the same size.
In this embodiment, the panoramic image is long, and the general display device can only display a part of the panoramic image, so that it is difficult to directly label birds on the panoramic image, so that the panoramic image is cut according to the size given by the user, so that the calculation efficiency of the algorithm can be ensured, valuable characteristic information of the image can be reserved as far as possible, the labeling efficiency is improved, and the labeling cost is reduced.
And S106, estimating the bird density of the undetermined image according to the depth density estimation neural network model, and obtaining a bird density map.
In this embodiment, a depth density estimation (Depth Density Estimation) neural network model is presented for estimating bird density, and the DDE model includes two parts: front-end feature extraction and rear-end density map generation, wherein the background of image data is complex, the image data contains information of various objects, a depth residual error network is adopted to extract the front-end features, and a bottomless residual error network extends the depth neural network to 152 layers through short-circuit connection. The method effectively improves the extraction of the neural network to the image characteristics, and solves the problems of degradation and difficult training and convergence along with the deep penetration of the neural network. The ResNet network is modified on the basis of the VGG network, and a residual error unit is added through a short circuit mechanism, so that the problem that the deep network is difficult to train is solved. Five structures with different depths exist, wherein the ResNet152 deepens the network structure based on the ResNet34, and the feature extraction capability is stronger. The method comprehensively considers the precision of the whole network and the execution efficiency of the algorithm, and selects the ResNet34 network with high calculation speed as a feature extraction part of the front end to be connected with a density map generation part of the rear end on the premise of losing a small amount of precision. The entire connection layer of ResNet is removed. In addition, resNet34 is pre-trained on the tens of millions of data sets ImageNet, with weights containing rich target feature information. Therefore, the network structure is used as a feature extraction network, so that the convergence rate of network training can be effectively increased.
In this embodiment, the dilation convolution increases the Receptive Field of the convolution operation by adding a hole calculation (hole calculation) operation to the convolution kernel, such that the output of each convolution contains a broad range of information without the need for a pooling operation (pooling operation). In a common convolution operation, the convolution kernel has a size of 3×3 and a receptive field of 3×3. In the dilation convolution, when a dilation ratio of 1 is used, the size of the convolution kernel is still 3×3, but the receptive field becomes 7×7. Because birds in panoramic images are smaller and there may be individual occlusions, richer, more complete local feature information is needed to generate the feature map. Thus, the back-end density map generation section is mainly constituted by hole convolution. In order to preserve the image feature information as much as possible, the pooling layer is not used.
In this embodiment, a plurality of dimension reduction operations are used by the res net34 in generating a density map, and the size of the obtained density map is several times that of the original image, so that an error exists between the generated density map and the density map label of the original image, and the error formula is as follows:
where M is the number of samples in the dataset, M, n is the image pixel coordinates, num GT Is the number of birds that are directly read from the tag file. When the number of downsampling exceeds four, errors generated by the tag greatly affect the accuracy of the algorithm. Although the downsampling is smaller than 4 times and the error is smaller, as the output characteristic diagram increases, the number of data required to be processed by the algorithm also increases, and the algorithm is not suitable for convergence. Thus, the two-dimensional reduction operation is deleted from the DDE. The size of the output feature map is 128×96, so that the convergence speed of the algorithm is ensured, and the accuracy of the algorithm is ensured.
And S108, splicing the bird density map into a panoramic bird density map according to a splicing function, and calculating the number of birds.
In this embodiment, the stitching function is a concat function, which can integrate a plurality of images with the same size, restore the images to a panoramic image, output a panoramic density image, and calculate the number of birds.
According to the bird density estimation method, the number and density of the birds can be accurately identified, counted and estimated by using an image flow technology and a deep neural network. The method provides abundant data support for ecological research and wetland protection, and reveals the distribution and utilization rules of the water birds in the wetland.
In one embodiment, as shown in fig. 2, the acquiring the bird panoramic image may specifically include the following steps:
step S202, a fixed-length video is acquired.
Step S204, extracting the key image frames of the fixed-length video according to a preset extraction sequence.
And step S206, splicing the key image frames to obtain a panoramic image.
In this embodiment, a video track is called from a randomly set tour through a PyCharm to obtain a video including information about the entire scene, key frames of the video are obtained by using OpenCV, the key frames are named according to numbers, the key frames are extracted according to a preset extraction sequence, an exemplary length of a section of panoramic shot video is four seconds, each second is composed of thirty frames of images, 120 images in total, and corresponding digital key frame images are extracted from the section of video according to the preset extraction sequence (19, 26, 33, 41, 49, 56, 63, 71, 79, 86, 92, 99, 106, 110, 116), and the extracted images are spliced to obtain a panoramic image.
In one embodiment, as shown in fig. 3, the splitting pretreatment is performed on the bird panoramic image to obtain a plurality of undetermined images with the same size, and specifically includes the following steps:
step S302, determining a minimum unit of image clipping;
step S304, adjusting the size of the panoramic image to be a multiple of the minimum unit;
and step S306, cutting the panoramic image to obtain a pending image.
In this embodiment, due to the uncertainty of the rotation parameters of the image capturing device, the acquired image data is not fixed in size, the length of the original image ranges from 4k to 30k, the width ranges from 1k to 1.2k, the image needs to be scaled, the pixel length and width of the image become several times of the length and width of the minimum unit, the minimum unit of image clipping is determined, the size of the minimum unit is 1024×768, an automatic clipping method is designed to obtain a plurality of minimum processing units, and the formula is as follows:
wherein N is the minimum number of processing units x, y is the true pixel value of the input image after clipping, x 0 Is a fixed length of 1024 pixels, y 0 Is 768 pixels high. The actual pixel value of the input image divided by 1024 (the number of fixed pixel values); the whole integer input image of actual pixel values is divided by 1024 (the number of still pixel values), recorded and replaced by the whole integer b. The actual pixel value of the input image m is divided by 1024 to obtain a fixed pixel value remainder lambda. For adjusting the size of the y-direction scale.
In this embodiment, the size of the image with the smallest unit is 1024×768, so that the calculation efficiency of the algorithm can be ensured, valuable characteristic information of the image can be kept as far as possible, the original image is cut into a plurality of small images with 1024×768 through cutting pretreatment of the image, the labeling efficiency of birds in the image is improved, each bird in the image is marked with a red dot, and coordinate points are stored for further counting.
In one embodiment, as shown in fig. 4, the establishment of the depth density estimation neural network model specifically includes the following steps:
step S402, a data set is established according to the bird panoramic image;
step S404, generating a density map label for the sample image in the data set according to an adaptive Gaussian kernel algorithm;
step S406, training the sample images in the data set according to a depth residual error network to obtain a density map;
and step S408, performing error calculation based on the density map and the density map label.
In this embodiment, as shown in fig. 6, a depth density estimation neural network is required to be established for bird density estimation on an acquired image, a large number of images are required to be used as training samples for neural network training, namely, a data set of bird images is required to be established through data collection, data filtering and data labeling, the data collection is based on the acquired bird panoramic image, and due to frequent habitat movement and weather change of birds, the complexity of data characteristics is very high, so that a data set suitable for neural network training is required to be acquired, a large number of data acquired by a camera are required to be filtered, and meanwhile, in order to acquire data sets of images with different scenes and different sizes, the parameter conditions of an image acquisition device are changed to acquire panoramic images of diversified focal lengths, angles and shooting scenes; the annotation tool for counting the bird population is used for annotating the data, a point is placed at the center of a target, the coordinates of the point are stored to represent a bird, the panoramic image is usually long, the size of a general display device is limited, only a partial area of the panoramic image can be displayed, and the annotation of the whole panoramic image is difficult, so that all samples in the data set are subjected to clipping pretreatment, the annotation efficiency is improved, the annotation cost is reduced, and the marked tag information can be mapped to the name and the image size in the original data through the corresponding relation among files through the three-step clipping pretreatment.
In this embodiment, the density map may show the spatial distribution of birds in a given image relative to the total number of birds, and an adaptive Gaussian blur algorithm is used to generate a density map label, which essentially consists of two parts, namely an image annotation display and an image conversion representation, first, a two-dimensional matrix with accurate resolution based on the image is generated, then the coordinates are transformed by input-output resolution comparison, and then the transformed label coordinates x are transformed by a delta function i Set to 1, as follows:
the two-dimensional gaussian kernel is then convolved with the unit pulse function. However, x of different samples i And are not completely independent, and have perspective distortion relation. Therefore, the problem of perspective distortion must be considered in processing. Calculating the current sample point x by using a K nearest neighbor algorithm i And its surrounding sample point x i+1 When (1):
wherein d is i Is each x i The average distance corresponding to the sample, β, is the scaling parameter. When β=0.5, a Depth Density Estimation (DDE) neural network model is proposed for bird density estimation, the DDE comprising two parts: front-end feature extraction and rear-end density map generation, wherein the background of image data is complex, information of various objects is contained, the front-end feature extraction is carried out by using a depth residual error network with the best effect in the image recognition field, the ResNet network is modified on the basis of the VGG network, a residual error unit is added through a short circuit mechanism, and the problem that the depth network is difficult to train is solved. Five structures with different depths exist, wherein the ResNet152 deepens the network structure based on the ResNet34, and the feature extraction capability is stronger. The ResNet34 network with high calculation speed is selected as a feature extraction part of the front end by comprehensively considering the precision of the whole network and the execution efficiency of the algorithm, and is connected with a density map generation part of the rear end. And the entire connection layer of the res net is removed.
In this embodiment, the dilation convolution increases the Receptive Field of the convolution operation by adding an hole calculation operation to the convolution kernel, such that the output of each convolution contains a wide range of information without the need for a pooling operation (pooling operation). In a common convolution operation, the convolution kernel has a size of 3×3 and a receptive field of 3×3. In the dilation convolution, when a dilation ratio of 1 is used, the size of the convolution kernel is still 3×3, but the receptive field becomes 7×7. Because birds in the dataset are smaller and there may be individual occlusions, richer, more complete local feature information is needed to generate the feature map. Thus, hole convolution is selected as the main component of the back-end density map generation section. In order to preserve the image feature information as much as possible, we do not use a pooling layer.
In this embodiment, because ResNet34 uses multiple dimension-reduction operations in developing a density map, the size of the map is 1024 times the original image size; it is necessary to adjust the size of tag data. Experiments show that after the direct use adjustment operation, errors exist among labels, and an error calculation formula is as follows:
where M is the number of samples in the dataset, M, n is the image pixel coordinates, num GT Is the number of birds that are directly read from the tag file. When the number of downsampling exceeds four, errors generated by the tag greatly affect the accuracy of the algorithm. Although the downsampling is smaller than 4 times and the error is smaller, as the output characteristic diagram increases, the number of data required to be processed by the algorithm also increases, and the algorithm is not suitable for convergence. Thus, the two-dimensional reduction operation is deleted from the DDE. The size of the output characteristic diagram is 128 multiplied by 96, so that the convergence speed of the algorithm is ensured, and the accuracy of the algorithm is ensured.
In this example, 80% of the images in the dataset were used for training of the DDE model, the remaining 20% were used for DDE model testing and validation, and the initial model specifications were set as follows: the learning rate is 10 multiplied by 10 -5 Batch size 1, momentum 0.95, weight decay 5×10 -3 The iteration period is 200 epochs.
The validity of the DDE model was verified by Mean Absolute Error (MAE) and Root Mean Square Error (RMSE):
wherein N is the number of images to be detected, zi is the group trunk of the ith image (artificially marked by 20 workers and representing the number of real birds in the picture), and Zl is model prediction. MAE indicates the accuracy of the algorithm, while MSE evaluates the robustness and stability of the algorithm. To further express the accuracy of the algorithm directly, we calculate the accuracy of the algorithm as:
MAE and Error ratio The lower the algorithm accuracy on the test set, the lower the MSE, the more stable the algorithm and the better the adaptability. In addition, model MEA, RMSE and accuracy were calculated using test data sets of different sample amounts to illustrate the effectiveness of the method.
For example, one dataset covers different monitoring head scenes and has rich habitat characteristic information, the dataset contains 935 panoramic images with different sizes, manual statistics show that 787,552 birds are in the dataset, calculated MAE and RMSE are 120.86 and 599.74 respectively, meanwhile, the average error rate is 14.14%, the image processing rate is 2.12 frames per second, the average recognition accuracy of the images is 85.59%, the accuracy reaches 90.92% of the highest when the number of birds in the images is 800-1000, and when the number of birds in the images exceeds 2000, the accuracy is reduced, and the depth density estimation method has good adaptability to scenes with different sizes, different target densities and aggregation modes.
In one embodiment, as shown in fig. 5, the method for creating a dataset according to a panoramic image of birds specifically includes the following steps:
step S502, screening the panoramic image according to the resolution, definition and data statistics characteristics of the panoramic image to obtain a data set;
and step S504, labeling the panoramic image in the data set.
In this embodiment, in order to improve the quality of the data set, a large amount of acquired data is filtered, and screening is mainly performed in three aspects, and because the image features are not obvious, the image stitching is incomplete, and part of data is lost, the image with the resolution lower than 4k in the data set is removed; because the images acquired by the camera in the rotation process are affected by motion blur, the unclear photo sample in the data set needs to be removed; in order to ensure the rationality of data set distribution and the effectiveness of neural network training, the sample numbers of images with different bird numbers in the data set are required to be reasonably distributed, the images with less than 10 targets in the data set are removed, and the images with the bird numbers of 50-20000 are screened to ensure the reasonable distribution of the data set.
According to the method, the frame of automatic snapshot, splicing and statistics counting is provided by combining the monitoring video and the deep learning animal identification and counting, good accuracy is achieved, the overall error rate is 14.14%, and in addition, the DDE model with ResNet34 has a good effect on images with diversified bird densities and aggregation modes, so that the potential of computer vision in the aspects of wild animal monitoring and protection can be fully exerted. Compared with the traditional crowd density estimation method, the method has higher image resolution, the image is refined by cutting the large image, the training and reasoning capacity of the model and the attention of the model to local spatial information of the original image are improved, the trained model can effectively predict the number of the water birds appearing in the image and generate the density image, so that high frequency number estimation is realized, and a standardized and automatic method is provided for quickly and accurately monitoring the regional population change trend. In addition, the method can also support local management personnel to develop an automatic monitoring and displaying system, so that the system has the capability of remotely grasping and timely early warning the bird population change in the supervision area in real time, and particularly in remote areas which cannot be patrol regularly and areas with insufficient manpower patrol funds.
As shown in fig. 7, in one embodiment, there is provided a water bird density estimating apparatus, the apparatus comprising:
an image processing module 110 for acquiring a bird panoramic image; performing segmentation pretreatment on the bird panoramic image to obtain a plurality of undetermined images with the same size;
a bird density estimating module 120, configured to estimate bird density of the undetermined image according to a depth density estimating neural network model, to obtain a bird density map;
and the splicing module 130 is used for splicing the bird density map into a panoramic bird density map according to a splicing function and calculating the number of birds.
In this embodiment, the specific working method flow and technical effects of each module of the bird density estimation device are described in the foregoing, and are not described herein again.
FIG. 10 illustrates an internal block diagram of a computer device in one embodiment. As shown in fig. 10, the computer device includes a processor, a memory, a network interface, an input device, and a display screen connected by a system bus. The memory includes a nonvolatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program which, when executed by a processor, causes the processor to implement a method of aquatic bird density estimation. The internal memory may also have stored therein a computer program which, when executed by the processor, causes the processor to perform a method of aquatic bird density estimation. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, the input device of the computer equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by those skilled in the art that the structure shown in fig. 10 is merely a block diagram of some of the structures associated with the present application and is not limiting of the computer device to which the present application may be applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a bird density estimation device provided herein may be implemented in the form of a computer program that is executable on a computer device as shown in fig. 10. The memory of the computer device may store various program modules constituting the apparatus, such as the image processing module 110, the bird density estimating module 120, and the splicing module 130 shown in fig. 9. The computer program of each program module causes the processor to carry out the steps of a method for estimating bird density according to each embodiment of the present application described in the present specification.
In one embodiment, a computer device is presented, the computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program: acquiring a bird panoramic image;
performing segmentation pretreatment on the bird panoramic image to obtain a plurality of undetermined images with the same size;
estimating bird density of the undetermined image according to the depth density estimation neural network model to obtain a bird density map;
and splicing the bird density map into a panoramic bird density map according to a splicing function, and calculating the number of birds.
In one embodiment, a computer readable storage medium is provided, having a computer program stored thereon, which when executed by a processor causes the processor to perform the steps of:
acquiring a bird panoramic image;
performing segmentation pretreatment on the bird panoramic image to obtain a plurality of undetermined images with the same size;
estimating bird density of the undetermined image according to the depth density estimation neural network model to obtain a bird density map;
and splicing the bird density map into a panoramic bird density map according to a splicing function, and calculating the number of birds.
It should be understood that, although the steps in the flowcharts of the embodiments of the present application are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in various embodiments may include multiple sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor do the order in which the sub-steps or stages are performed necessarily performed in sequence, but may be performed alternately or alternately with at least a portion of the sub-steps or stages of other steps or other steps.
Those skilled in the art will appreciate that all or part of the processes in the methods of the above embodiments may be implemented by a computer program for instructing relevant hardware, where the program may be stored in a non-volatile computer readable storage medium, and where the program, when executed, may include processes in the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above-described embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above-described embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples only represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the present application. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (9)
1. A method of aquatic bird density estimation, the method comprising:
acquiring a bird panoramic image;
performing segmentation pretreatment on the bird panoramic image to obtain a plurality of undetermined images with the same size;
estimating bird density of the undetermined image according to the depth density estimation neural network model to obtain a bird density map;
and splicing the bird density map into a panoramic bird density map according to a splicing function, and calculating the number of birds.
2. The method of claim 1, wherein the acquiring the bird panoramic image comprises the steps of:
acquiring a fixed-length video;
extracting key image frames of the fixed-length video according to a preset extraction sequence;
and splicing the key image frames to obtain the panoramic image.
3. The method for estimating the bird density according to claim 1, wherein the splitting pretreatment is performed on the bird panoramic image to obtain a plurality of undetermined images with the same size, and the method comprises the following steps:
determining a minimum unit of image clipping;
adjusting the panoramic image size to be a multiple of the minimum unit;
and cutting the panoramic image to obtain a pending image.
4. A method of aquatic bird density estimation according to claim 3, wherein the cropping image comprises the steps of:
obtaining a cropped undetermined image by the following formula:
wherein N is the minimum unit number obtained by cutting; x, y is the length and width of the true pixel value of the input image, x is the fixed length of the minimum unit pixel, y is the fixed width of the minimum unit pixel, and λ is the width-direction scaling size.
5. The method for estimating the density of a bird according to claim 1, wherein the establishment of the depth density estimating neural network model comprises the steps of:
establishing a data set according to the bird panoramic image;
generating a density map label for the sample image in the dataset according to a self-adaptive Gaussian blur algorithm;
training the sample images in the data set according to a depth residual error network to obtain a density map;
and performing error calculation based on the density map and the density map label.
6. The method of claim 5, wherein the creating a dataset from the bird panoramic image comprises the steps of:
screening the panoramic image according to the resolution, definition and data statistics characteristics of the panoramic image to obtain a data set;
the data statistics characteristic is the number of birds in the panoramic image;
and labeling the panoramic image in the dataset.
7. A water bird density estimation device, the device comprising:
the image processing module is used for acquiring the bird panoramic image; performing segmentation pretreatment on the bird panoramic image to obtain a plurality of undetermined images with the same size;
the bird density estimation module is used for estimating the bird density of the undetermined image according to the depth density estimation neural network model to obtain a bird density map;
and the splicing module is used for splicing the bird density images into a panoramic bird density image according to a splicing function and calculating the number of birds.
8. A computer device comprising a memory and a processor, the memory having stored therein a computer program which, when executed by the processor, causes the processor to perform the steps of a water bird density estimation method according to any one of claims 1 to 6.
9. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, causes the processor to perform the steps of a waterfowl density estimation method according to any of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311668122.3A CN117670892A (en) | 2023-12-07 | 2023-12-07 | Aquatic bird density estimation method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311668122.3A CN117670892A (en) | 2023-12-07 | 2023-12-07 | Aquatic bird density estimation method and device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117670892A true CN117670892A (en) | 2024-03-08 |
Family
ID=90069372
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311668122.3A Pending CN117670892A (en) | 2023-12-07 | 2023-12-07 | Aquatic bird density estimation method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117670892A (en) |
-
2023
- 2023-12-07 CN CN202311668122.3A patent/CN117670892A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Song et al. | Spatiotemporal satellite image fusion using deep convolutional neural networks | |
| Basalamah et al. | Scale driven convolutional neural network model for people counting and localization in crowd scenes | |
| Dujon et al. | Machine learning to detect marine animals in UAV imagery: Effect of morphology, spacing, behaviour and habitat | |
| CN111046880A (en) | Infrared target image segmentation method and system, electronic device and storage medium | |
| CN111354024B (en) | Behavior prediction method of key target, AI server and storage medium | |
| CN111210399B (en) | Imaging quality evaluation method, device and equipment | |
| CN112329702B (en) | Method and device for rapid face density prediction and face detection, electronic equipment and storage medium | |
| Rong et al. | Pest identification and counting of yellow plate in field based on improved mask r-cnn | |
| CN109492665A (en) | Detection method, device and the electronic equipment of growth period duration of rice | |
| CN113192646A (en) | Target detection model construction method and different target distance monitoring method and device | |
| CN111091093A (en) | Method, system and related device for estimating number of high-density crowds | |
| CN113706481A (en) | Sperm quality detection method, sperm quality detection device, computer equipment and storage medium | |
| Guo et al. | Pigeon cleaning behavior detection algorithm based on light-weight network | |
| Song et al. | Remote sensing image spatiotemporal fusion via a generative adversarial network with one prior image pair | |
| CN112883915A (en) | Automatic wheat ear identification method and system based on transfer learning | |
| Zou et al. | Density estimation method of mature wheat based on point cloud segmentation and clustering | |
| Song et al. | Benchmarking wild bird detection in complex forest scenes | |
| Azhar et al. | A framework for multiscale intertidal sandflat mapping: A case study in the Whangateau estuary | |
| CN117670892A (en) | Aquatic bird density estimation method and device, computer equipment and storage medium | |
| CN115240168A (en) | Perception result obtaining method and device, computer equipment and storage medium | |
| Bai et al. | Video target detection of East Asian migratory locust based on the MOG2-YOLOv4 network | |
| Kathirvel et al. | Systematic number plate detection using improved yolov5 detector | |
| CN116935310A (en) | Real-time video monitoring bird density estimation method and system based on deep learning | |
| Ahmed et al. | Satellite image future landscape prediction using conditional adversarial networks | |
| Hežeľová | Artificial neural network for precipitation nowcasting |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |